Consumers' load profile determination based on different classification methods

Abstract

The restructuring of the electric power sector toward a fully competitive market gives an important role to the load profiles representing consumers' load-consumption pattern. They are obtained from the field measurements of individual consumers' load curves, and can be divided into two approaches. The first is based on the predefined consumers classes, the second uses pattern recognition methods to derive typical load profiles (TLP) from the obtained measurements. Since, it is clear that no single approach for classification is "optimal", multiple methods have to be used to verify the obtained results. For that purpose the hierarchic clustering algorithms and fuzzy c-means algorithm are applied. Results obtained demonstrate the ability of the used algorithms to classify different daily load curves and to generate comparable results. The most similar results of applied clustering algorithms were obtained by fuzzy c-means algorithm and hierarchical clustering algorithm with Ward distance between clusters.